Real time leak detection system applied to oil pipelines using sonic technology and neural networks

This work proposes a leak detection system using sonic technology, wavelet transform and neural networks to decompose and analyze pressure signals from oil pipelines in real time. The similarity between pressure and sound signals makes it possible to treat the first through digital filtering and wavelet decomposition together with a neural network to characterize and classify leak profiles. The leak detection system logic is embedded on 32 bit/150 MHz floating point DSPs. This system uses piezoresistive sensors, converters to the communication interface (Ethernet) and GPS devices, which are responsible for synchronizing reports and leak alarms. The DSPs code was written using ANSI C language.